File: ds_read2_superreg.ll

package info (click to toggle)
llvm-toolchain-17 1%3A17.0.6-22
  • links: PTS, VCS
  • area: main
  • in suites: forky, sid, trixie
  • size: 1,799,624 kB
  • sloc: cpp: 6,428,607; ansic: 1,383,196; asm: 793,408; python: 223,504; objc: 75,364; f90: 60,502; lisp: 33,869; pascal: 15,282; sh: 9,684; perl: 7,453; ml: 4,937; awk: 3,523; makefile: 2,889; javascript: 2,149; xml: 888; fortran: 619; cs: 573
file content (209 lines) | stat: -rw-r--r-- 10,495 bytes parent folder | download | duplicates (4)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
; RUN: llc -march=amdgcn -mcpu=bonaire -verify-machineinstrs -mattr=+load-store-opt,-enable-ds128 < %s | FileCheck --check-prefix=CI %s

@lds = addrspace(3) global [512 x float] undef, align 4
@lds.v2 = addrspace(3) global [512 x <2 x float>] undef, align 4
@lds.v3 = addrspace(3) global [512 x <3 x float>] undef, align 4
@lds.v4 = addrspace(3) global [512 x <4 x float>] undef, align 4
@lds.v8 = addrspace(3) global [512 x <8 x float>] undef, align 4
@lds.v16 = addrspace(3) global [512 x <16 x float>] undef, align 4

; CI-LABEL: {{^}}simple_read2_v2f32_superreg_align4:
; CI: ds_read2_b32 [[RESULT:v\[[0-9]+:[0-9]+\]]], v{{[0-9]+}} offset1:1{{$}}
; CI: s_waitcnt lgkmcnt(0)
; CI: buffer_store_dwordx2 [[RESULT]]
; CI: s_endpgm
define amdgpu_kernel void @simple_read2_v2f32_superreg_align4(ptr addrspace(1) %out) #0 {
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds  [512 x <2 x float>], ptr addrspace(3) @lds.v2, i32 0, i32 %x.i
  %val0 = load <2 x float>, ptr addrspace(3) %arrayidx0, align 4
  %out.gep = getelementptr inbounds <2 x float>, ptr addrspace(1) %out, i32 %x.i
  store <2 x float> %val0, ptr addrspace(1) %out.gep
  ret void
}

; CI-LABEL: {{^}}simple_read2_v2f32_superreg:
; CI: ds_read_b64 [[RESULT:v\[[0-9]+:[0-9]+\]]], v{{[0-9]+}}{{$}}
; CI: s_waitcnt lgkmcnt(0)
; CI: buffer_store_dwordx2 [[RESULT]]
; CI: s_endpgm
define amdgpu_kernel void @simple_read2_v2f32_superreg(ptr addrspace(1) %out) #0 {
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x <2 x float>], ptr addrspace(3) @lds.v2, i32 0, i32 %x.i
  %val0 = load <2 x float>, ptr addrspace(3) %arrayidx0
  %out.gep = getelementptr inbounds <2 x float>, ptr addrspace(1) %out, i32 %x.i
  store <2 x float> %val0, ptr addrspace(1) %out.gep
  ret void
}

; CI-LABEL: {{^}}simple_read2_v4f32_superreg_align4:
; CI-DAG: ds_read2_b32 v[[[REG_X:[0-9]+]]:[[REG_Y:[0-9]+]]], v{{[0-9]+}} offset1:1{{$}}
; CI-DAG: ds_read2_b32 v[[[REG_Z:[0-9]+]]:[[REG_W:[0-9]+]]], v{{[0-9]+}} offset0:2 offset1:3{{$}}
; CI-DAG: v_add_f32_e32 v[[ADD0:[0-9]+]], v[[REG_X]], v[[REG_Z]]
; CI-DAG: v_add_f32_e32 v[[ADD1:[0-9]+]], v[[REG_Y]], v[[REG_W]]
; CI: v_add_f32_e32 v[[ADD2:[0-9]+]], v[[ADD0]], v[[ADD1]]
; CI: buffer_store_dword v[[ADD2]]
; CI: s_endpgm
define amdgpu_kernel void @simple_read2_v4f32_superreg_align4(ptr addrspace(1) %out) #0 {
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x <4 x float>], ptr addrspace(3) @lds.v4, i32 0, i32 %x.i
  %val0 = load <4 x float>, ptr addrspace(3) %arrayidx0, align 4
  %elt0 = extractelement <4 x float> %val0, i32 0
  %elt1 = extractelement <4 x float> %val0, i32 1
  %elt2 = extractelement <4 x float> %val0, i32 2
  %elt3 = extractelement <4 x float> %val0, i32 3

  %add0 = fadd float %elt0, %elt2
  %add1 = fadd float %elt1, %elt3
  %add2 = fadd float %add0, %add1

  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %x.i
  store float %add2, ptr addrspace(1) %out.gep
  ret void
}

; CI-LABEL: {{^}}simple_read2_v3f32_superreg_align4:
; CI-DAG: ds_read2_b32 v[[[REG_X:[0-9]+]]:[[REG_Y:[0-9]+]]], v{{[0-9]+}} offset1:1{{$}}
; CI-DAG: ds_read_b32 v[[REG_Z:[0-9]+]], v{{[0-9]+}} offset:8{{$}}
; CI-DAG: v_add_f32_e32 v[[ADD0:[0-9]+]], v[[REG_X]], v[[REG_Z]]
; CI-DAG: v_add_f32_e32 v[[ADD1:[0-9]+]], v[[ADD0]], v[[REG_Y]]
; CI: buffer_store_dword v[[ADD1]]
; CI: s_endpgm
define amdgpu_kernel void @simple_read2_v3f32_superreg_align4(ptr addrspace(1) %out) #0 {
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x <3 x float>], ptr addrspace(3) @lds.v3, i32 0, i32 %x.i
  %val0 = load <3 x float>, ptr addrspace(3) %arrayidx0, align 4
  %elt0 = extractelement <3 x float> %val0, i32 0
  %elt1 = extractelement <3 x float> %val0, i32 1
  %elt2 = extractelement <3 x float> %val0, i32 2

  %add0 = fadd float %elt0, %elt2
  %add1 = fadd float %add0, %elt1

  %out.gep = getelementptr inbounds float, ptr addrspace(1) %out, i32 %x.i
  store float %add1, ptr addrspace(1) %out.gep
  ret void
}

; CI-LABEL: {{^}}simple_read2_v4f32_superreg_align8:
; CI: ds_read2_b64 [[REG_ZW:v\[[0-9]+:[0-9]+\]]], v{{[0-9]+}} offset1:1{{$}}
; CI: buffer_store_dwordx4 [[REG_ZW]]
; CI: s_endpgm
define amdgpu_kernel void @simple_read2_v4f32_superreg_align8(ptr addrspace(1) %out) #0 {
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x <4 x float>], ptr addrspace(3) @lds.v4, i32 0, i32 %x.i
  %val0 = load <4 x float>, ptr addrspace(3) %arrayidx0, align 8
  %out.gep = getelementptr inbounds <4 x float>, ptr addrspace(1) %out, i32 %x.i
  store <4 x float> %val0, ptr addrspace(1) %out.gep
  ret void
}

; CI-LABEL: {{^}}simple_read2_v4f32_superreg:
; CI-DAG: ds_read2_b64 [[REG_ZW:v\[[0-9]+:[0-9]+\]]], v{{[0-9]+}} offset1:1{{$}}
; CI: buffer_store_dwordx4 [[REG_ZW]]
; CI: s_endpgm
define amdgpu_kernel void @simple_read2_v4f32_superreg(ptr addrspace(1) %out) #0 {
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x <4 x float>], ptr addrspace(3) @lds.v4, i32 0, i32 %x.i
  %val0 = load <4 x float>, ptr addrspace(3) %arrayidx0
  %out.gep = getelementptr inbounds <4 x float>, ptr addrspace(1) %out, i32 %x.i
  store <4 x float> %val0, ptr addrspace(1) %out.gep
  ret void
}

; FIXME: Extra moves shuffling superregister
; CI-LABEL: {{^}}simple_read2_v8f32_superreg:
; CI-DAG: ds_read2_b64 [[VEC_HI:v\[[0-9]+:[0-9]+\]]], v{{[0-9]+}} offset0:2 offset1:3{{$}}
; CI-DAG: ds_read2_b64 [[VEC_LO:v\[[0-9]+:[0-9]+\]]], v{{[0-9]+}} offset1:1{{$}}
; CI-DAG: buffer_store_dwordx4 [[VEC_HI]], v[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0 addr64 offset:16
; CI-DAG: buffer_store_dwordx4 [[VEC_LO]], v[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0 addr64{{$}}
; CI: s_endpgm
define amdgpu_kernel void @simple_read2_v8f32_superreg(ptr addrspace(1) %out) #0 {
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x <8 x float>], ptr addrspace(3) @lds.v8, i32 0, i32 %x.i
  %val0 = load <8 x float>, ptr addrspace(3) %arrayidx0
  %out.gep = getelementptr inbounds <8 x float>, ptr addrspace(1) %out, i32 %x.i
  store <8 x float> %val0, ptr addrspace(1) %out.gep
  ret void
}

; FIXME: Extra moves shuffling superregister
; CI-LABEL: {{^}}simple_read2_v16f32_superreg:
; CI-DAG: ds_read2_b64 [[VEC0_3:v\[[0-9]+:[0-9]+\]]], v{{[0-9]+}} offset1:1{{$}}
; CI-DAG: ds_read2_b64 [[VEC4_7:v\[[0-9]+:[0-9]+\]]], v{{[0-9]+}} offset0:2 offset1:3{{$}}
; CI-DAG: ds_read2_b64 [[VEC8_11:v\[[0-9]+:[0-9]+\]]], v{{[0-9]+}} offset0:4 offset1:5{{$}}
; CI-DAG: ds_read2_b64 [[VEC12_15:v\[[0-9]+:[0-9]+\]]], v{{[0-9]+}} offset0:6 offset1:7{{$}}
; CI: s_waitcnt lgkmcnt(0)
; CI-DAG: buffer_store_dwordx4 [[VEC0_3]], v[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0 addr64{{$}}
; CI-DAG: buffer_store_dwordx4 [[VEC4_7]], v[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0 addr64 offset:16
; CI-DAG: buffer_store_dwordx4 [[VEC8_11]], v[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0 addr64 offset:32
; CI-DAG: buffer_store_dwordx4 [[VEC12_15]], v[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0 addr64 offset:48
; CI: s_endpgm
define amdgpu_kernel void @simple_read2_v16f32_superreg(ptr addrspace(1) %out) #0 {
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x <16 x float>], ptr addrspace(3) @lds.v16, i32 0, i32 %x.i
  %val0 = load <16 x float>, ptr addrspace(3) %arrayidx0
  %out.gep = getelementptr inbounds <16 x float>, ptr addrspace(1) %out, i32 %x.i
  store <16 x float> %val0, ptr addrspace(1) %out.gep
  ret void
}

; Do scalar loads into the super register we need.
; CI-LABEL: {{^}}simple_read2_v2f32_superreg_scalar_loads_align4:
; CI-DAG: ds_read2_b32 v[[[REG_ELT0:[0-9]+]]:[[REG_ELT1:[0-9]+]]], v{{[0-9]+}} offset1:1{{$}}
; CI-NOT: v_mov {{v[0-9]+}}, {{[sv][0-9]+}}
; CI: buffer_store_dwordx2 v[[[REG_ELT0]]:[[REG_ELT1]]]
; CI: s_endpgm
define amdgpu_kernel void @simple_read2_v2f32_superreg_scalar_loads_align4(ptr addrspace(1) %out) #0 {
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %x.i
  %arrayidx1 = getelementptr inbounds float, ptr addrspace(3) %arrayidx0, i32 1

  %val0 = load float, ptr addrspace(3) %arrayidx0
  %val1 = load float, ptr addrspace(3) %arrayidx1

  %vec.0 = insertelement <2 x float> undef, float %val0, i32 0
  %vec.1 = insertelement <2 x float> %vec.0, float %val1, i32 1

  %out.gep = getelementptr inbounds <2 x float>, ptr addrspace(1) %out, i32 %x.i
  store <2 x float> %vec.1, ptr addrspace(1) %out.gep
  ret void
}

; Do scalar loads into the super register we need.
; CI-LABEL: {{^}}simple_read2_v4f32_superreg_scalar_loads_align4:
; CI-DAG: ds_read2_b32 v[[[REG_ELT0:[0-9]+]]:[[REG_ELT1:[0-9]+]]], v{{[0-9]+}} offset1:1{{$}}
; CI-DAG: ds_read2_b32 v[[[REG_ELT2:[0-9]+]]:[[REG_ELT3:[0-9]+]]], v{{[0-9]+}} offset0:2 offset1:3{{$}}
; CI-NOT: v_mov {{v[0-9]+}}, {{[sv][0-9]+}}
; CI: buffer_store_dwordx4 v[[[REG_ELT0]]:[[REG_ELT3]]]
; CI: s_endpgm
define amdgpu_kernel void @simple_read2_v4f32_superreg_scalar_loads_align4(ptr addrspace(1) %out) #0 {
  %x.i = tail call i32 @llvm.amdgcn.workitem.id.x() #1
  %arrayidx0 = getelementptr inbounds [512 x float], ptr addrspace(3) @lds, i32 0, i32 %x.i
  %arrayidx1 = getelementptr inbounds float, ptr addrspace(3) %arrayidx0, i32 1
  %arrayidx2 = getelementptr inbounds float, ptr addrspace(3) %arrayidx0, i32 2
  %arrayidx3 = getelementptr inbounds float, ptr addrspace(3) %arrayidx0, i32 3

  %val0 = load float, ptr addrspace(3) %arrayidx0
  %val1 = load float, ptr addrspace(3) %arrayidx1
  %val2 = load float, ptr addrspace(3) %arrayidx2
  %val3 = load float, ptr addrspace(3) %arrayidx3

  %vec.0 = insertelement <4 x float> undef, float %val0, i32 0
  %vec.1 = insertelement <4 x float> %vec.0, float %val1, i32 1
  %vec.2 = insertelement <4 x float> %vec.1, float %val2, i32 2
  %vec.3 = insertelement <4 x float> %vec.2, float %val3, i32 3

  %out.gep = getelementptr inbounds <4 x float>, ptr addrspace(1) %out, i32 %x.i
  store <4 x float> %vec.3, ptr addrspace(1) %out.gep
  ret void
}

; Function Attrs: nounwind readnone
declare i32 @llvm.amdgcn.workitem.id.x() #1

; Function Attrs: nounwind readnone
declare i32 @llvm.amdgcn.workitem.id.y() #1

attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }
attributes #2 = { convergent nounwind }